Procedure for Creating a Virtual Multibank Agent: Using Intelligent Agents
Then, in the second phase of our study, we will search for a procedure that enables “discriminating” the information obtained and selecting that which best adapts to a preestablished objective. When there are many variables, before designing any method of classification, it is necessary to select among the original variables those that are really relevant; that is, the problem has to do with finding a subset of variables with which it is possible to carry out the task of optimum classification. Thus, for instance, when searching for information about loans, the user would want to find the most amount of information in the shortest amount of time possible and with the minimum amount of navigation sessions (an objective that can be achieved by providing the virtual multibank agent with a tabu algorithm in its design), but the user could also obtain a classification of the offers based on his/her aspirations about the product or service searched. In order to achieve this objective, we will use an intelligent agent that works without human supervision and will offer the user the result of an automatic exploration for which a visual map of the different banking offers on a certain product will be obtained, arranged according to their similarity.
Intelligent agents are programs that enable making searches using self-organizing maps in order to establish categories of terms that are used for defining or representing vectors of the web pages that will be used for establishing a visual map of the same pages placed according to similarity. Once one of the pages has been selected, the user can use this map to obtain those that are closer or more related to it. review
For the search through the previous categorization of pages, we will start with a database that contains the collection of web pages corresponding to the financial institutions that operate in Spain and that offer their products and services online. In the searches using previous categorization, such classification is usually carried out manually, although there are interesting experiences of automatic classification of web pages, such as for example the WEBSOM project, which uses self-organizing maps in order to establish categories of terms. These categories are used for defining or representing vectors of the websites, with which it is possible to construct a visual map of the same pages placed according to similarity, by applying the same mechanism. Once one of the pages has been selected, the user can use this map to obtain those that are closer or more related to it.
In our study, we will design a virtual multibank intelligent agent that is able to provide its services to a Spanish-speaking user. The possibility of being able to expand the search method to gather multilingual data will be the subject of future research.
The objective that we propose is that the intelligent agents are able to be designed for semi-automatically processing information located on bank websites, searching for information related with that found on the pages visited by the user and with the search requirements established by the user. In sum, the intelligent agent designed will be in charge of searching for links between resources, comparing and processing Internet information for the user who requests such information, and unlike a simple search engine, it could even interact with the environment, purchasing a financial service or product, if the user wishes to do so. In which case, the agent would use protocols that enable it to work with the interfaces of these financial entities.